Weighting
RTI statisticians have done considerable research to reduce survey bias from nonresponse and coverage errors. Our researchers deploy a variety of innovative techniques from logistic modeling of unit response propensity to our unique generalized exponential modeling (GEM) procedures, as well as simpler techniques, such as weighting class adjustments.
Our GEM procedures (now available in SUDAAN®) are similar to logistic models using bounds for adjustment factors and bounds on variance inflation. The GEM approach provides a unified method for the nonresponse adjustment and poststratification. It has a built-in extreme weight control mechanism through applying different bounds to the predetermined extreme weights. Therefore, the extreme weights can be kept in control while the weights are being adjusted.
Focus Areas
- Estimation of survey bias
- Reduction of nonresponse and coverage bias
Methodologies/Techniques
- Iterative proportional fitting (or raking) adjustments
- Poststratification
- Calibration
- Truncation and smoothing of extreme weights
- Weighting class adjustments
- Propensity modeling
- Adjustments for unknown eligibility
rti.org